Experiment Setup¶

Algorithm¶

- Scheduling time T: 1 (in block-time e.g once per day if block arrival is once per day)
- Scheduling policy: FCFS
- No budget unlocking: data-lifetime=0.1, n=1 (At time 0.1 block gets fully unlocked)       

Workload¶

- Total number of blocks: 600 (initial blocks=1)
- One task per block-unit
- Blocks arrival time: 1 (e.g. 1 day)
- Number of blocks: all previous blocks
- Tasks lifetime: 1 (block-time e.g. 1 day = after that the task can't run)
- Budget demand (epsilon): 1 (fixed for all incoming tasks for now)
- Block budget initial capacity: 10
- Number of tasks: (See task arrival time below)
- Number of distinct queries: 1
- Query types: Count only
- RDP

Cost Model for dynamic programming planner as of now¶

- Minimizing aggregations + binary structure constraint
In [1]:
from notebooks.caching.utils import analyze_experiment
tasks_path = "~/privacypacking/data/covid19/covid19_workload/privacy_tasks.csv"
In [2]:
# analyze_experiment(tasks_path, "run_and_report_2022-12-04_22-14-38/")
In [3]:
# analyze_experiment(tasks_path, "run_and_report_2022-12-04_23-02-19/")
In [4]:
# analyze_experiment(tasks_path, "exp-1000/")
In [5]:
# analyze_experiment(tasks_path, "2objexp-500/")
In [6]:
# analyze_experiment(tasks_path, "exp-500/")
In [7]:
# analyze_experiment(tasks_path, "2objexp1/")
In [8]:
analyze_experiment(tasks_path, "errorobj/")
In [9]:
analyze_experiment(tasks_path, "finalexp")
In [ ]: